Accepted Manuscript

Improved Radial Heat Sink For Led Lamp Cooling

Vítor A.F. Costa , António M.G. Lopes

PII: S1359-4311(14)00332-9
DOI: 10.1016/j.applthermaleng.2014.04.068
Reference: ATE 5592

To appear in: Applied Thermal Engineering

Received Date: 9 July 2013

Accepted Date: 25 April 2014

Please cite this article as: V.A.F. Costa, A.M.G. Lopes, Improved Radial Heat Sink For Led Lamp
Cooling, Applied Thermal Engineering (2014), doi: 10.1016/j.applthermaleng.2014.04.068.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT
IMPROVED RADIAL HEAT SINK FOR LED LAMP COOLING

Vítor A. F. Costa1, António M. G. Lopes2

1 - Departamento de Engenharia Mecânica, Universidade de Aveiro, Campus Universitário

de Santiago, 3810-193 Aveiro, Portugal

PT
2 - ADAI-LAETA, Departamento de Engenharia Mecânica, Universidade de Coimbra 3030-
788 Coimbra, Portugal

RI
Abstract

SC
This paper presents a numerical study concerning an improved heat sink for a light emitting
diodes (LED) lamp operating under natural convection conditions. Basic geometry of the heat sink
is of cylindrical nature, to be obtained from cutting a aluminium extruded bar comprising a

U
cylindrical central core and a number of uniformly distributed radial fins. Minimum diameter of the
AN
central core is fixed and the parameters to be explored are the number of fins, their thickness, length
(radial dimension) and height. Although not included in the numerical simulations, the thermal
M

resistance due to the use of a thin thermal interface material (TIM) layer between the LED lamp
back and the heat sink is taken into account in the analysis. The main objective of the heat sink is to
D

cool the LED lamp so that the lamp maximum temperature at the contact region with the heat sink
is maintained below the critical temperature given by the manufacturer. This is a crucial aspect in
TE

what concerns the expected lifetime of the LED lamp and should be achieved at the expenses of a as
low as possible aluminium mass. Taking these criteria in mind, a design procedure is proposed and
EP

followed in the search for the improved heat sink to cool a particular LED lamp. Results obtained
with the commercial code ANSYS-CFX clearly show the relative importance of the different
C

governing parameters on the heat sink performance and allow the choice of the better solution
within the frame of dimensional constrains. Although the present results concern a particular LED
AC

lamp, the proposed methodology can be extended to other types of heat sinks for general light
and/or electronic components.

Keywords: LED lamps, heat sink, thermal performance, natural convection heat transfer, conjugate
heat transfer

Nomenclature
cP constant pressure specific heat [J/(kg.ºC)] g gravitational acceleration [m/s2]
1
 ACCEPTED MANUSCRIPT
H fin height [m]

Hc height of calculation domain [m] Greek symbols

i number of project variable [-] α thermal diffusivity [m2/s]

k thermal conductivity [W/(m.ºC)] β volumetric expansion coefficient [1/K]

L fin length, including central core [m] ∆ difference value

Lc radius of calculation domain [m] δ thickness of the TIM layer [m]

PT
M heat sink mass [kg] θ slice angle for calculation domain [rad]

RI
n number of fins [-] ρ density [kg/m3]
p pressure [Pa] µ dynamic viscosity [kg/(m.s)]

SC
Pr Prandtl number [-] ν kinematic viscosity [m2/s]
Q& heat flow [W]

r central core radius [m]

U
AN
R Thermal resistance [ºC/W] Subscripts

Ra Rayleigh number [-] c core

M

t fin thickness [m] i optimization variable index

D

T temperature [ºC] r reference

TE

u, v, w Cartesian velocity components [m/s] s solid

x, y, z Cartesian coordinates [m] ∞ ambient

EP

Xi Generic project variable

1 - INTRODUCTION
C

In the past few years, LED lamps have been assuming a large role in the illumination market,
AC

mainly due to their potential in creating not only light but interesting light environments, associated
with low power consumption even when compared with other energy-saving lamp types. Still, from
the energetic viewpoint, LED lamps have not yet the desirable efficiency, as a considerable amount
of energy is released as heat. This tends to increase the temperature of the LED lamps, leading to a
decrease of their lifetime. In order to meet the expected long lifetime, LED lamps must operate
below a certain temperature threshold, as given by the manufacturer. To ensure this, the heat sink
associated to the LED lamps must provide the needed cooling, requiring the minimum mass of the
involved material to obtain the heat sink. Additionally, internal electronic circuitry may have to be

2
 ACCEPTED MANUSCRIPT
activated by cutting the electrical energy to the LED lamp, changing the illumination
characteristics, if by any reason some overshoot on the temperature of the LED lamp occurs.

Thus, one may realize that the design of heat sinks used to cool LED lamps is of major
importance to ensure their long lifetime. Heat sinks design criteria should contemplate not only
their thermal performance, but also the associated manufacturing costs, which are directly related to
the total material mass and production processes involved. For the latest criterion, a uniform cross

PT
section is desirable as it can be obtained from simple bars obtained by extrusion, followed by a
cutting operation.

RI
Heat sinks for LED lamps can operate under natural or forced convection, the first being the
preferred ones as neither additional fans nor electric consumption are required. Additionally, noise

SC
generation by electric motors and fans is also avoided if natural convection heat sinks are adopted.
On the other hand, natural convection heat sinks are thermally less effective, being consequently

U
less compact and heavier, and need a careful design to reach the required thermal performance.
AN
Heat sinks and LED lamps are assembled together using a thin layer of TIM, which, although
designed to have a high thermal conductivity, locally increases the thermal resistance of the
M

assembly.

Some studies can be found in the literature concerning the cooling of LED lamps. This can
D

include what happens inside [1,2] or outside the LED module using detailed CFD studies in a
TE

steady-state approach, such as in [3-12], some of them including the packing aspects leading to the
LED module. More specifically, Luo et al. [1] conducted a numerical and experimental study
considering the whole set composed by the vapour chamber and a finned heat sink, and Arika et al.
EP

[2] consider the thermal management of the set composed by an assembly of LED lamps and a
finned heat sink. In the work by Ying at al. [3], the optimization of the heat sink associated to a
C

high power LED spot lamp is numerically studied. Scheepers and Visser [4] studied the heat
AC

management of high power LEDS using heat sinks through a numerical approach, and comparison
is made between the detailed thermal model and simpler thermal resistance models. In the work by
Christensen and Graham [5], a 3D numerical simulation is presented for an array of high power
LED lamps together with a heat sink, and the thermal resistance network is analysed trying to
estimate the different contributions for the heat management, in the search of compact LED
systems. Chi et al. [6] performed the thermal analysis of high power LED lamps and the associated
heat sink, using a CFD numerical simulation together with some heat transfer correlations,
including radiation heat transfer. By its own turn, in the study by Weng [7] it is shown how a

3
 ACCEPTED MANUSCRIPT
detailed 3D CFD analysis can improve the thermal performance of LED illumination systems. The
study by Yu et al. [8] deals with the numerical simulation and optimization of a radial heat sink,
with two alternative configurations having a void central part, for LED cooling purposes,
concluding that it is impossible to optimize both thermal performance and heat sink mass, the work
by Yu et al. [9] including also some experimental results for a similar heat sink configuration.
Huang et al. [10] studied the thermal dynamics of the overall set of a LED fixture including the
luminaire, the LED lamps and the heat sink, and Houl et al. [13] studied the thermal dynamics of a

PT
LED array system wit in line pin fin heat sink. Ha [11] considered the numerical simulation of a
high power LED package and extracted the values of the most relevant thermal resistances of the

RI
system. Yu et al. [12] analysed the effect of radiation heat transfer on the thermal performance of
radial heat sinks similar to that considered in [8] and [9]. Studies considering the problem as a

SC
simple thermal resistance combination can also be found [14,15]. Some other studies also include
the set composed by the LED module, the heat sink and the luminaries, like in [16]. The recent

U
study by Yu et al. [12] shows how radiation heat transfer is relevant when analyzing the thermal
behaviour of this kind of heat sinks. Agostini et al. [16] present a detailed state of the art for high
AN
heat flux cooling technologies. In the work by Shyu et al. [17], a 270x1W LED array together with
a plate fin heat sink in an acrylic housing is experimentally studied. The complete CFD numerical
M

simulation of single-phase active liquid cooling systems, including minichannels for liquid
circulation, can be found in [18]. Huang et al. [19] very recently proposed a constant power driving
D

control for a 150 W LED luminaire, thus stabilizing the illumination of LED under large
temperature variations.
TE

However, in spite of the referred studies, no work was found concerning the heat sinks for the
EP

recent 4th generation LED lamps, like the FORTIMO lamps produced by Philips, more specifically
when using a radial heat sink with a central core, obtained by simple extrusion and cutting
manufacturing processes. Even if the heat sink by itself is not an energy saving device, it is crucial
C

for the right operation during the expected long lifetime of the energy saving devices that are the
AC

LED lamps, and is a relevant applied thermal engineering problem.

The main objective of the present work is to find the improved geometrical configuration for an
aluminium heat sink to be used in 4th generation FORTIMO LED lamps, given the heat release
rate, critical (maximum) temperature of the LED module and space constraints. The improved
configuration is, in the present context, that corresponding to the lowest heat sink mass that
guarantees the required cooling level. Any economic analysis, even if requiring reliable information
which may be difficult to obtain, will lead coincidently to the cheaper heat sink, obtained from the
lowest mass of material and requiring the simplest manufacturing processes to obtain it. The
4
 ACCEPTED MANUSCRIPT
procedure proposed in the present work leads to an improvement path that, starting from an initial
guess, is governed by the partial derivatives of the maximum (core) temperature with respect to the
heat sink mass, for each geometrical parameter under study. This is different from an integrated
optimization procedure, as no special optimization techniques and tools are needed, and the
improved heat sink can be obtained by non-experts in optimization studies, following a way with its
own physical insight and not as an optimization mathematical problem. Some of the results
obtained with the commercial code ANSYS-CFX during the improvement process are presented,

PT
along with the final solution corresponding to the founded improved effective heat sink.

RI
2 - PHYSICAL MODELING

As the heat sink is considered in the vertical position, as presented in Fig. 1, relatively cold ambient

SC
air approaches the hot heat sink and becomes warmer, thus rising and pulling new fresh air from the
colder ambient to satisfy continuity. The resulting flow assumes the form of a rising plume, air

U
entering the heat sink in an essentially radial (horizontal) direction and leaving it in an essentially
AN
vertical direction. Heat sink remains warm due to the heat released by the LED lamps assembly, the
manufacturer of the LED lamp giving the value of the heat released to the heat sink, which needs to
be released by the heat sink to the natural convection air flow. The balance between the heat
M

released by the LED lamp to the heat sink and the heat released by the heat sink to the cooling air
dictates the temperature at the centre (core) of the LED lamp module, which needs to be below a
D

critical temperature value specified by the LED lamp manufacturer. More effective heat sinks lead
TE

to lower temperatures at the centre of the LED lamp module, the best heat sink being that allowing
a temperature at the center of the LED lamp module close to the critical value given by the LED
EP

lamp manufacturer but requiring the lowest mass of material to obtain it.

2.1 - Physical model

C

The heat sink to be studied is radial, with a circular base and a number of equally spaced
AC

rectangular fins joining at a central core. The heat sink seats on the top (back) of the LED module,
as depicted in Figure 1a). The recent FORTIMO DLM GEN 4 3000/830 LED module considered in
this work features a rectangular base releasing the maximum thermal power of 19.2 W through the
8.3×7.8 cm2 rectangular base. The manufacturer advises a maximum temperature at the LED base
below 70 ºC, when operating under an ambient air temperature of 35 ºC.

The adopted heat sink starting configuration for the improvement process (basic configuration)
has a base diameter similar to the LED module largest diagonal, thus ensuring full contact with the

5
 ACCEPTED MANUSCRIPT
heat sink at LED module base. Table 1 presents the physical dimensions according to the schematic
drawing in Figure 1b). The considered heat sink includes a thin aluminium disk between the back of
the LED module and the main body of the heat sink, mainly for fixing purposes. However, other
solutions including screws acting over the fins themselves can be designed for that purpose.

2.2 - Governing equations

Natural convection is governed by the Prandtl number:

PT
ν
Pr = (1)
α

RI
and by the Rayleigh number:

SC
g β ∆T Lr 3
Ra = Pr (2)
ν2

U
where all properties refer the air involving the heat sink.
AN
For the present case, taking the reference length Lr as the heat sink base diameter and the
representative temperature difference between the heat sink and ambient air as ∆T ≈ 40o C, Eq. (2)
M

computes Ra ≈ 6 × 106 , which is well below the typical transition value 108-109 for the turbulent
regime [20]. Therefore, laminar conditions are assumed, corresponding to a conjugate laminar
D

natural convection 3D problem in steady-state conditions.

TE

Warming up and cooling down periods of LED lamps are considerably short when compared
with the continuous longer and nearly steady operating periods, thus justifying that only the steady
EP

state operation conditions are considered in the present work. Governing equations are the mass
conservation equation, the momentum (Navier-Stokes) equations and the thermal energy
conservation equation, which read, respectively:
C

∂ ∂ ∂
( ρu ) + ( ρ v ) + ( ρ w) = 0
AC

(3)
∂x ∂y ∂z

∂ ∂ ∂ ∂p  ∂ 2u ∂ 2 u ∂ 2 u 
( ρ uu ) + ( ρ vu ) + ( ρ wu ) = − + µ  2 + 2 + 2  (4a)
∂x ∂y ∂z ∂x  ∂x ∂y ∂z 

∂ ∂ ∂ ∂p  ∂ 2v ∂ 2v ∂ 2v 
( ρ uv ) + ( ρ vv ) + ( ρ wv ) = − + µ  2 + 2 + 2  (4b)
∂x ∂y ∂z ∂y  ∂x ∂y ∂z 

∂ ∂ ∂ ∂p  ∂2w ∂2w ∂2w 

( ρ uw) + ( ρ vw) + ( ρ ww) = − + µ  2 + 2 + 2  − ρ g (4c)
∂x ∂y ∂z ∂z  ∂x ∂y ∂z 

6
 ACCEPTED MANUSCRIPT
∂ ∂ ∂ k  ∂ 2T ∂ 2T ∂ 2T 
( ρ uT ) + ( ρ vT ) + ( ρ wT ) =  2 + 2 + 2 

(1)(5)
∂x ∂y ∂z cP  ∂x ∂y ∂z 

where the work of pressure forces and viscous dissipation were neglected in the thermal energy
conservation equation. Within the solid, advection terms vanish, leading to the heat conduction
equation:

∂  ∂T  ∂  ∂T  ∂  ∂T 
0=  ks  +  ks  +  ks  (6)

PT
∂x  ∂x  ∂y  ∂y  ∂z  ∂z 

In the previous equations, fluid and solid properties were considered as constant, expect for fluid

RI
density in the buoyancy term on the z momentum equation, which is computed based on the Ideal
Gas Law.

SC
Due to the expected temperature differences between solid surfaces and the environment and
physical proximity between the solid surfaces, heat transfer by radiation must be taken into account.

U
It represents an additional heat transfer mode promoting heat transfer from the base of the heat sink
AN
to its fins, and thus increases the effectiveness of the heat sink. Several models exist to consider the
surface radiation heat transfer in the numerical model, with specific capabilities depending on the
M

physical problem under analysis. In the present case, due to the transparent nature of the air, the
Monte Carlo radiation model represents the best option [21]. This model simulates the interaction
D

between photons and the environment, a process that is carried out by following the path of a
predetermined number of photons, including surface reflection, absorption and scattering, and it is
TE

one of the radiation model options included in the ANSYS-CFX software package.
EP

2.3 - Computational model and boundary conditions

ANSYS-CFX software package is a well-known and well-established software package, which
C

has been used and validated under many different conditions, and also to solve natural convection
AC

problems. As a certified software, extensive validation has been done by different research groups
under a great variety of situations, with corresponding documentation available through the internet
and in different reports available from ANSYS, including both laminar and turbulent flows [22], as
well as free convection with conjugate heat transfer [23]. CFX uses a coupled solver in a co-located
(non-staggered) mesh. In order to avoid the well know checkerboard pressure field, a fourth-order
Rhie- Chow interpolation scheme is used. In the present simulations, advection terms were
modelled using the CFX-high resolution scheme option, which ensures boundness of the solution. A
trilinear interpolation scheme was adopted for the evaluation of the diffusion terms and pressure
gradient As based on a finite volume formulation, discretization equations, and in particular the
7
 ACCEPTED MANUSCRIPT
enthalpy (thermal energy) equations, are obtained for each control volume. Boundaries between
fluid and solid domains coincide with control volume faces at these locations, and no boundary
conditions need to be specified as a normal conservative condition is considered. Thus, the obtained
discretized equations for enthalpy ensure thermal energy conservation in the fluid-solid interface,
the conjugate heat transfer being automatically taken into account through this integration
approach.

PT
For the sake of simulations, the physical presence of the LED module was not taken into
account. This leads to a geometry that possesses a periodic angular behaviour, allowing the
simulations to be carried out for a single fin. Placement of computational domain boundaries and

RI
mesh size were carefully chosen so as to guarantee results’ independence on both mesh and
calculation domain. On the other hand, corresponding values such as number of mesh nodes and

SC
domain dimensions were tested in order to keep computer requirements at reasonable levels. Figure
2a shows the computational domain geometry adopted for the simulations, for which Lc = 0.13 m

U
and H c = 0.09 m. Different mesh sizes were tested as well, leading to a solution characterized by a
AN
total of about 200,000 nodes (cf. Figure 2b), using a target inner mesh size in the air domain of
0.002 m. In the air-solid interface, where typical mesh elements size is 0.0005 m, an inflated layer
M

2×10-5 m high with prismatic elements was employed so as to better simulate the heat transfer
mechanism therein.
D

For the solid domain, the heat flux value was imposed at the bottom boundary (heat sink base),
TE

so as to satisfy a total of 19.2 W (the heat flow released by the LED array). Lateral boundaries of
the calculation domain were assigned an angular periodic condition, while non-slip conditions were
EP

applied at all the solid-gas interfaces. For the gas domain, besides the angular periodicity applied to
the corresponding boundaries, an open-type condition with zero relative pressure and 35 ºC
temperature was considered for all the exterior boundaries. Inlet velocities at the different portions
C

of the external boundaries are thus obtained as that required by the natural convection problem
AC

under analysis. If part of an exterior boundary is an outlet for fluid a parabolic exit boundary
condition is considered there. The solid-gas interfaces are treated with an energy-conservative
condition, which is automatically taken into account as the numerical method is based on a control
volume approach. For radiation calculations, aluminium surfaces are assumed as gray with diffusive
reflection, featuring an emissivity of 0.85. A total of 10,000 photons was used for the application of
the Monte Carlo Model, considering a surrounding temperature of 35 ºC.

8
 ACCEPTED MANUSCRIPT
Obtained results qualitatively agree with results for similar LED lamp modules obtained from the
literature, and no exhaustive comparison can be made as no other numerical or experimental results
are known for the FORTIMO DLM GEN 4 3000/830 LED module.

2.4 - Thermal resistance at the LED module-heat sink contact

The attachment between the aluminium heat sink and the LED module is made with screws. In
order to reduce the thermal resistance between both solids, a TIM is commonly used, so that no air

PT
gaps due to surface roughness are present in the interface to considerably increase the contact
thermal resistance.

RI
The temperature drop at the contact interface may be easily computed using a simple 1D

SC
analysis. The governing heat conduction equation is:

∆T Q& δ
Q& = k Ac i ⇒ ∆Ti = (7)
δ

U
AN k Ac

The thermal conductivity of a good TIM material, like DOW CORNING TC-5022, is typically 4
W/(m ºC). Considering a film thickness of 0.0002 m, the total heat release rate, and heat sink base

area as presented previously, an interface temperature drop ∆Ti = 0.077 o C is obtained from Eq. (7).
M

Therefore, one may conclude that the additional temperature drop due to the contact resistance can
D

be neglected in subsequent analysis. However, if this temperature drop is not so low, and if it is to
be considered, the one dimensional analysis is followed through the thin TIM layer and the
TE

numerical domain refers only to the solid aluminium body of the heat sink, the maximum
admissible temperature at the heat sink being the maximum temperature admissible at the LED
EP

module minus the temperature drop through the TIM layer. Even so, to accommodate any
uncertainty on the calculations, as well as the negative effects on cooling of some dirty that can
C

accumulate over the surfaces of the heat sink, a maximum core temperature of 65 ºC was taken as
the objective in this study.
AC

3 - IMPROVEMENT STRATEGY
As previously pointed out, the objective of the present work is to obtain the improved heat sink
geometrical configuration so as to guarantee that the LED module maximum temperature value
(verified at the LED module base) is below the manufacturer threshold. This is to be achieved with
a as low as possible heat sink mass, provided the geometrical dimensions are kept within reasonable
limits. It is anticipated that an increase on the thickness of the fins, and increase on the length of the
fins, and an increase on the height of the heat sink will result on an increased thermal effectiveness
9
 ACCEPTED MANUSCRIPT
of the heat sink. It is also anticipated that an increase on the number of fins will result in an increase
on the thermal effectiveness of the heat sink below a given number of fins, but that this behaviour is
inverted above that given number of fins, as a high number of fins results in a smaller and narrower
available space for the natural convection risen flow. However, the increase of any of the referred
parameters will result on an increase of the mass needed to obtain the heat sink, and it needs to be
evaluated what is the parameter to be increased resulting in the higher increase on the thermal
effectiveness of the heat and simultaneously to the lower increase on the mass needed to obtain the

PT
heat sink. This is the criterion to be followed when searching for an improved heat sink to cool the
particular LED lamps assembly considered, starting from a given non-optimal initial heat sink

RI
geometry.

The geometry parameterization was set according to the project variables listed in Table 1: fin

SC
height H, fin thickness t, fin length L and number of fins, n. The core minimum radius r1 , also listed
in this table, is kept constant. The maximum temperature Tc (core temperature) to be evaluated at

U
the base of the heat sink (as referred, due to its small value, the temperature drop across the TIM
AN
material is neglected), is a function of these variables:

Tc = Tc ( X i ) , i = 1, 2,3, 4; ( X i = H , t , L, n ) (8)
M

A change ∆ X i of one of the project variables X i will lead to a change in both the core
D

temperature ∆Tci and the heat sink mass ∆M i :

TE

∂Tci ∂M i
∆Tci = ∆ X i ; ∆M i = ∆Xi (9)
∂X i ∂X i
EP

It can be easily seen that, attending to the definition of each project variable, increasing any of
the project variables will lead to a mass increase − any of the partial mass derivative in the previous
equation is always positive and can be analytically calculated. On the other hand, the core
C

temperature partial derivatives must be evaluated through simulation and their sign is most often
AC

negative (except, as will be seen, for ∂T / ∂n at large values of n). Therefore, one may realize that,
since the improvement path to follow should be evaluated in terms of both mass and core
temperature variations, the governing variable is ∂Tci / ∂M i , which is negative in the improvement

space − decreasing the core temperature is achieved at the expenses of an increase of the heat sink
mass.

The improvement path criterion will thus be:

10
 ACCEPTED MANUSCRIPT
o Improvement variables i with a large value ∂Tci / ∂M i should be changed towards a mass

increase, in order to achieve a “large” core temperature decrease at the expenses of “small”
mass increase ;

o Improvement variables i with a small value ∂Tci / ∂M i should be changed towards a mass

decrease, in order to decrease heat sink mass at the expenses of only a small core
temperature increase.

PT
Based on the previous considerations, the improvement strategy is:

RI
o Starting with the basic configuration (cf. Table 1), evaluate the absolute value of the partial
derivatives ∂Tci / ∂M i through numerical simulation, by changing, in turn, each

SC
improvement parameter and using the obtained values for Tc and M readily available in the
software post-processor.

U
o Identify the improvement parameter(s) with the largest ∂Tci / ∂M i value and proceed the
AN
improvement towards lower Tc values, taking also into account geometry constraints. In this
sense, the number of fins n will be a priority as this improvement parameter doesn´t
M

influence the total occupied volume; on the other hand, fin length L and height H should be
kept at the admissible minimum values.
D

o Fine tune the improvement, by checking the heat sink dimensions and decreasing the heat
TE

sink mass using improvement variable(s) with low ∂Tci / ∂M i .

EP

4 - RESULTS AND ANALYSIS

The starting point for the improvement process is the basic configuration detailed in Table 1, for
C

which numerical computations led to a core temperature Tc = 87.65 o C and a heat sink mass
AC

M = 0.197 kg. For the heat sink thermal resistance based on the maximum core temperature, given
by

Tc − T∞
R= (10)
Q&i

a value of 2.742 ºC/W was obtained. These results are synthesized in Table 2. Figure 3 depicts a
visualization of the results for: (a) The velocity field at a vertical plane located between two fins;
and (b) The temperature field at the heat sink. The difference between the heat sink temperature and

11
 ACCEPTED MANUSCRIPT
the ambient air temperature creates buoyancy forces that produce an ascending air movement
towards the heat sink core. As the air mass passes between the fins, its temperature increases,
creating a higher temperature region in the plume central part, as it is well illustrated by the
temperature coloured vectors in Figure 3. The temperature range at the solid domain is only about
0.8 ºC, due to the considerably high thermal conductivity of the aluminium; even so, as expected,
temperatures at the heat sink core are higher than those at the exposed part of the fin.

PT
Table 3 lists the adopted data for the initial variation of the improvement parameters, necessary
for the evaluation of the partial derivatives of core temperature with respect to the mass of the heat
sink. As expected, fin length is the parameter with the largest influence in the core temperature.

RI
Although these results would point out a fin length increase as the best choice to proceed to
improve the heat sink, the total occupied volume would also increase by a large amount, since for

SC
the basic configuration the fins already protrude outside the heat sink base. As pointed out
previously, the number of fins should be a priority parameter since the total occupied volume is not

U
affected when changing this parameter. Dependence of the core temperature and ∂Tci / ∂M i with
AN
this parameter is shown in Figure 4. It is interesting to observe the evolution of these variables with
n, revealing that the effectiveness of adding more fins vanishes as the number of fins increases. This
M

is a consequence of the small spacing between each two adjacent fins, which prevents fresh air to
penetrate the inter-fin space and remove heat from there, decreasing the overall heat sink
D

effectiveness. It is, therefore, not advisable to choose a too large number of fins.
TE

Based on these results, a maximum of 40 fins was established, corresponding to a core

temperature Tc = 72.85 o C and to a heat sink mass M = 0.316 kg.
EP

Analysing the partial derivatives presented in Table 3, one may note that a fin thickness change
will allow reducing mass without a significant core temperature increase. Accordingly, as a next
C

step, fin thickness was reduced from 2 mm (basic configuration) to 1 mm, from which a core
AC

temperature Tc = 73.47 o C and a heat sink mass M = 0.197 kg were obtained, for n = 40 fins. It is
interesting to check that, with this reduced fin thickness, the dependence of Tc with n became larger
and the inversion of the ∂Tci / ∂M i sign was not found up to n = 70, as shown in Figure 5. By
increasing the number of fins and reducing their thickness, one could reduce the core temperature
by approximately 14 ºC and the thermal resistance by 27%, for a total mass similar to the initial
configuration.

12
 ACCEPTED MANUSCRIPT
Further reduction from Tc = 73.47 o C to values below 70 ºC can be achieved by increasing the
fin height. Simulation results are as depicted in Figure 6 and in Table 4. From a simple linear
interpolation, one could anticipate a value H = 0.028 m necessary for Tc = 65 o C, which was

confirmed with the numerical simulation, corresponding to a heat sink mass M = 0.244 kg. The final
configuration for the heat sink is depicted in Figure 7, along with the visualization of the obtained
temperature field at the heat sink surface.

PT
One may then conclude that, starting from an original heat sink configuration with mass 0.197kg
leading to a core temperature Tc = 87.65 o C and thermal resistance R = 2.742 ºC/W, the proposed

RI
improvement allowed a reduction of 22.65 o C in the core temperature to reach Tc = 65 o C, for a

final thermal resistance R = 1.281 ºC/W. This was achieved at the expenses of a mass increase of

SC
only 24%, corresponding to a final mass M = 0.244 kg. It must be retained that this is the improved
solution, which satisfactorily provides the required cooling for the considered LED lamp involving

U
a satisfactory mass of aluminium, but that it is not the optimum solution in a strict sense, and that
AN
even slightly better solutions can be obtained through the finest tuning of the governing parameters,
following the described improvement strategy.
M

The results presented and discussed in the present work are specific for the FORTIMO DLM
GEN 4 3000/830 LED module. However, attending that improved or even optimum cooling
D

systems may be needed for other LED lamp modules, the used methodology, with its own physical
TE

insight and without requiring knowledge and use of optimization tools and software, can and must
be used when searching for the best cooling system for other LED lamp modules.
EP

5 - CONCLUSIONS
An improvement procedure was presented and followed to find the improved geometrical
C

configuration for a heat sink operating under natural convection conditions, to be used for the
AC

cooling of a recent FORTIMO 4th generation LED lamp. The objective was to achieve a maximum
core temperature of 65 ºC keeping the heat sink total mass and occupied volume contained.
Numerical simulations were done for conjugate laminar steady state natural convection, including
heat transfer by radiation through the Monte Carlo model. The improvement path is given by the
partial derivatives of the core temperature with respect to the mass of the heat sink, subjected to
occupied volume constraints. Improvement variables considered were number of fins, fin height, fin
length and fin thickness. Starting from a basic configuration leading to a core temperature 22.65 o C

above the required limit, the objective core temperature of 65 o C was reached by increasing the
13
 ACCEPTED MANUSCRIPT
number of fins and their height and reducing the fin thickness. These geometrical modifications
were responsible for a heat sink mass increase of only 0.047 kg, which represents approximately
24% of the initial mass, but to a considerable temperature decrease, up to reach the admissible
maximum core temperature of the LED module.

The presented methodology is general in character and may be extended to improve other types
of heat sinks for virtually any light and/or electronic components.

PT
6 - REFERENCES
1. X. Luo, R. Hu, T. Guo, X. Zhu, W. Chen, Z. Mao, S. Liu, Low thermal resistance LED light

RI
source with vapor chamber coupled fin heat sink, IEEE 2010 Electronic Components and
Technology Conference, Paris, Las Vegas, 2010 (doi: 10.1109/ECTC.2010.5490645).

SC
2. M. Arika, C. Beckerb, S. Weaverb, J. Petroskic, Thermal management of LEDs: package to
system, Third International Conference on Solid State Lighting SPIE, Bellingham, WA, 2004,
Vol. 5187 (doi: 10.1117/12.512731).
U
AN
3. Y. G.-Ying, Z. X. Ping, H. S.-Hong, H. W-Wen, G. T.-Tai, Thermal simulation and
optimization design on a high power LED spot lamp, Optoelectronics Letters 7 (2) (2011) 117-
M

121 (doi: 10.1007/s11801-011-0142-8).

4. G. Scheepers, J. A. Visser, Detailed thermal modeling of high powered LEDs, 25th IEEE SEMI-
D

THERM Symposium 2009, San José, California, 2009 (doi: 10.1109/STHERM.2009.4810747).

TE

5. A. Christensen, S. Graham, Thermal effects in packaging high power light emitting diode
arrays, Applied Thermal Engineering 29 (2009) 364–371 (doi:
EP

10.1016/j.applthermaleng.2008.03.019).

6. W.-H. Chi, T.-L. Chou, C.-N. Han, K.-N. Chiang, Analysis of thermal performance of high
C

power light emitting diodes package, IEEE 2008 10th Electronics Packaging Technology
Conference, Singapore, 2008 (doi: 10.1109/EPTC.2008.4763488).
AC

7. C.-J. Weng, Advanced thermal enhancement and management of LED packages, International
Communications in Heat and Mass Transfer 36 (2009) 245–248 (doi:
10.1016/j.icheatmasstransfer.2008.11.015).

8. S.-H. Yu, K.-S. Lee, S.-J. Yook, Optimum design of a radial heat sink under natural convection,
International Journal of Heat and Mass Transfer 54 (2011) 2499–2505 (doi:
10.1016/j.ijheatmasstransfer.2011.02.012).

14
 ACCEPTED MANUSCRIPT
9. S.-H. Yu, K.-S. Lee, S.-J. Yook, Natural convection around a radial heat sink, International
Journal of Heat and Mass Transfer 53 (2010) 2935–2938 (doi:
10.1016/j.ijheatmasstransfer.2010.02.032).

10. B.-J. Huang, C.-W. Tang, M.-S. Wu, System dynamics model of high-power LED luminaire,
Applied Thermal Engineering 29 (2009) 609–616 (doi: 10.1016/j.applthermaleng.2008.03.038).

11. M. S. Ha, Thermal Analysis of High Power LED Arrays, MSc. Thesis, School of Mechanical

PT
Engineering, Georgia Institute of Technology, Atlanta, GA, 2009.

12. S.-H. Yu, D. Jang, K.-S. Lee, Effect of radiation in a radial heat sink under natural convection,

RI
International Journal of Heat and Mass Transfer 55 (2012) 505–509 (doi:
10.1016/j.ijheatmasstransfer.2011.09.026).

SC
13. F. Houl, D. Yangl, G. Q. Zhangl, Y. Hail, D. Liul, L. Liul, Thermal transient analysis of LED
array system with in-line pin fin heat sink, 12th. Int. Conf. on Thermal, Mechanical and

U
Multiphysics Simulation and Experiments in Microelectronics and Microsystems, IEEE
EuroSimE 2011, Linz, Austria, 2011 (doi: 10.1109/ESIME.2011.5765844).
AN
14. X. Luo, W. Xiong, S. Liu, A simplified thermal resistance network model for high power LED
street lamp, IEEE International Conference on Electronic Packaging Technology & High
M

Density Packaging (ICEPT-HDP 2008), Shangai, China, 2008 (doi:

10.1109/ICEPT.2008.4606976).
D

15. X. Luo, W. Xiong, T. Cheng, S. Liu, Design and optimization of horizontally-located plate fin
TE

heat sink for high power LED street lamps, 2009 IEEE Electronic Components and Technology
Conference (2009) 854-859, San Diego, California, 2009 (doi: 10.1109/ECTC.2009.5074112).
EP

16. B. Agostini, M. Fabbri, J. E. Park, L. Wojtan, J. R. Thome, B. Michel, State of the art of high
heat flux cooling technologies, Heat Transfer Engineering 28(4) (2007) 258–281 (doi:
C

10.1080/01457630601117799).
AC

17. B. J.-C. Shyu, K.-W. Hsu, K.-S. Yang, C.-C. Wang, Thermal characterization of shrouded plate
fin array on an LED backlight panel, Applied Thermal Engineering 31 (2011) 2909-2915 (doi:
10.1016/j.applthermaleng.2011.05.019).

18. B. Ramos-Alvarado, B. Feng, G. P. Peterson, Comparison and optimization of single-phase

liquid cooling devices for the heat dissipation of high-power LED arrays, Applied Thermal
Engineering 59 (2013) 648-659 (doi: 10.1016/j.applthermaleng.2013.06.036).

15
 ACCEPTED MANUSCRIPT
19. B.-J. Huang, C.-W. Chen, C.-D. Ong, B.-H. Du, P.-C. Hsu, Development of constant-power
driving control for light-emitting-diode (LED) luminaire, Applied Thermal Engineering 50
(2013) 645-651 (doi: 10.1016/j.applthermaleng.2012.07.030).

20. A. Bejan, Heat Transfer, Wiley, New York, 1993.

21. M. Modest, Radiation Heat Transfer, McGraw-Hill, New York, 1993.

22. W. Vieser, T. Esch, F. Menter, Heat transfer predictions using advanced two-equation

PT
turbulence models, ANSYS CFX Validation Report, CFX-VAL10/0404, ANSYS Europe Ltd.,
2002.

RI
23. H. Xu, C. Guetari, K. Svihla, Simulation of free convection with conjugate heat transfer, 2006
International ANSYS Conference: A World of Simulation, Pittsburgh, PA, 2006.

U SC
AN
M
D
TE
C EP
AC

16
 ACCEPTED MANUSCRIPT

FIGURE CAPTIONS

Figure 1. The heat sink. (a) 3D view of the ensemble; (b) geometrical parameters.
Figure 2. Calculation domain, for a single fin, basic configuration. (a) Domain dimensions; (b)
Non-structured mesh.
Figure 3. Results for the basic configuration. (a) Velocity field vectors at the vertical plane
between two fins (vectors are coloured according to local temperature); (b) Temperature

PT
field at the heat sink surface.
Figure 4. Influence of number of fins, for t = 2 mm. (a) Core temperature versus number of fins;
(b) ∂Tc/∂M versus number of fins.

RI
Figure 5. Influence of number of fins, for t = 1 mm. (a) Core temperature versus number of fins;
(b) ∂Tc/∂M versus number of fins.

SC
Figure 6. Influence of fin height. (a) Core temperature versus fin height; (b) Core temperature
versus heat sink mass.
Figure 7. Final configuration for the improved heat sink. (a) Heat sink geometry; (b) Temperature

U
distribution at the heat sink surface.
AN
M
D
TE
C EP
AC

17
 ACCEPTED MANUSCRIPT
TABLES

Table 1. Basic geometric configuration of the heat sink.

Core radius Fin height Fin thickness Fin length N of fins/ Angle
r1 H t L n/θ

PT
0.01 m 0.021 m 0.002 m 0.063 m 20/18º

RI
SC
Table 2. Results for basic configuration

Core temperature Thermal Resistance Heat sink mass

U
Tc [ºC] R [ºC /W] M [kg]
AN
87.65 ºC 2.742 0.197
M
D

Table 3. Variation of optimization parameters for the initial calculation of partial derivatives.
TE

Fin height Fin thickness Fin length Number of fins

H t L n
EP

∆Xi 0.004 m -0.001 m -0.0053 m 5

∂Tci / ∂M i -260 ºC/kg -20 ºC/kg -800 ºC/kg -150 ºC/kg

C
AC

18
 ACCEPTED MANUSCRIPT
Table 4. Simulation results for changing fin height.

Thermal
Fin height Core temperature Heat sink mass
Resistance
H [m] Tc [ºC] M [kg]
R [ºC /W]
0.021 73.5 2.005 0.197
0.025 68.0 1.719 0.223
0.030 63.6 1.49 0.256
0.035 59.6 1.281 0.289

PT
RI
U SC
AN
M
D
TE
C EP
AC

19
 ACCEPTED MANUSCRIPT
FIGURES

PT
RI
SC
r1
L

U t
AN
M

z H
D
TE

(a) (b)
C EP
AC

Figure 1. The heat sink. (a) 3D view of the ensemble; (b) geometrical parameters.

20
 ACCEPTED MANUSCRIPT

PT
RI
Lc

U SC
AN
Hc
M
D

(a) (b)
TE
C EP
AC

Figure 2. Calculation domain, for a single fin, basic configuration. (a) domain dimensions. (b) non-
structured mesh.

21
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M

(a) (b)
D
TE
C EP
AC

Figure 3. Results for the basic configuration. (a) Velocity field vectors at the vertical plane between two
fins. Vectors are coloured according to local temperature; (b) Temperature field at the heat sink surface.

22
 ACCEPTED MANUSCRIPT

PT
RI
SC
90 150
Tc ∂Tc
∂M

U
0
AN
75

-150
M

60 -300
15 35 55 75 n 15 35 55 75 n
D

(a) (b)
TE
C EP
AC

Figure 4. Influence of number of fins, for t = 2 mm. (a) Core temperature versus number of fins; (b) ∂Tc/∂M
versus number of fins.

23
 ACCEPTED MANUSCRIPT

PT
RI
90 150

SC
Tc ∂Tc
∂M
0

U
75
AN
-150

60 -300
M

15 35 55 75 n 15 35 55 75 n
(a) (b)
D
TE
C EP
AC

Figure 5. Influence of number of fins, for t = 1 mm. (a) Core temperature versus number of fins; (b) ∂Tc/∂M
versus number of fins.

24
 ACCEPTED MANUSCRIPT

PT
RI
SC
75 75
Tc Tc

U
70 AN 70

65 65

60 60
M

55 55
0.02 0.025 0.03 0.035 0.04 H 0.15 0.20 0.25 0.30 M
D

(a) (b)
TE
C EP
AC

Figure 6. Influence of fin height. (a) Core temperature versus fin height; (b) Core temperature versus heat
sink mass.

25
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE

(a) (b)
C EP
AC

Figure 7. Final configuration for the improved heat sink. (a) Heat sink geometry; (b) Temperature
distribution at the heat sink surface.

26
 ACCEPTED MANUSCRIPT

IMPROVED RADIAL HEAT SINK FOR LED LAMP COOLING

Vítor A. F. Costa1, António M. G. Lopes2

1 - Departamento de Engenharia Mecânica, Universidade de Aveiro, Campus

Universitário de Santiago, 3810-193 Aveiro, Portugal
2 - ADAI-LAETA, Departamento de Engenharia Mecânica, Universidade de Coimbra

PT
3030-788 Coimbra, Portugal

HIGHLIGHTS FOR REVIEW

RI
A numerical study is made concerning the radial heat sink for a specific LED lamp.

SC
The cylindrical heat sink is obtained from an extruded aluminum bar.
The required cooling effect is obtained using the minimum mass of material.

U
AN
M
D
TE
C EP
AC